· Contributors · Organizations ·
Hypoglycemia Detection Using Hand Tremors: A Home Study in T1D Adults
Virtual Program Session
DescriptionDiabetes currently affects millions of people worldwide. Monitoring blood glucose (BG) is usually performed by invasive or intrusive methods, and their accuracy is still an open question. Hand tremor is a significant symptom of hypoglycemia as nerves and muscles are powered by blood sugar. There are no established techniques or algorithms for monitoring and detecting hypoglycemia through hand tremors. This paper proposes a non-invasive method to objectively detect hypoglycemia events through hand tremors using acceleration data. Different machine learning methods were utilized on the data collected from 33 Type 1 Diabetes patients to classify and differentiate between hypoglycemia (HG) and non-hypoglycemia (non-HG) states. The results obtained showed an accuracy of 81.46% and a recall of 78.59% with Ensemble Learning based on HG levels obtained from CGM-based readings. These results indicate that the proposed method can be a potential tool as an alert mechanism to detect hypoglycemia events.